Intradomain Allosteric Network Modulates Calcium Affinity of the C

Aug 25, 2016 - ADVERTISEMENT .... Intradomain Allosteric Network Modulates Calcium Affinity of the C-Type Lectin Receptor Langerin ... Calcium-Indepen...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/JACS

Intradomain Allosteric Network Modulates Calcium Affinity of the C‑Type Lectin Receptor Langerin Jonas Hanske,†,‡,⊥ Stevan Aleksić,‡,⊥ Martin Ballaschk,‡,§,# Marcel Jurk,∥ Elena Shanina,†,‡ Monika Beerbaum,§ Peter Schmieder,§ Bettina G. Keller,*,‡ and Christoph Rademacher*,†,‡ †

Department of Biomolecular Systems, Max Planck Institute of Colloids and Interfaces, Am Mühlenberg 1, 14424 Potsdam, Germany Institute of Chemistry and Biochemistry, Department of Biology, Chemistry, and Pharmacy, Freie Universität Berlin, Takustraße 3, 14195 Berlin, Germany § Leibniz-Institut für Molekulare Pharmakologie, Robert-Rössle-Straße 10, 13125 Berlin, Germany ∥ Department of Bioinformatics, Max Planck Institute for Molecular Genetics, Ihnestraße 63-73, 14195 Berlin, Germany ‡

S Supporting Information *

ABSTRACT: Antigen uptake and processing by innate immune cells is crucial to initiate the immune response. Therein, the endocytic C-type lectin receptors serve as pattern recognition receptors, detecting pathogens by their glycan structures. Herein, we studied the carbohydrate recognition domain of Langerin, a C-type lectin receptor involved in the host defense against viruses such as HIV and influenza as well as bacteria and fungi. Using a combination of nuclear magnetic resonance and molecular dynamics simulations, we unraveled the molecular determinants underlying cargo capture and release encoded in the receptor architecture. Our findings revealed receptor dynamics over several time scales associated with binding and release of the essential cofactor Ca2+ controlled by the coupled motions of two loops. Applying mutual information theory and site-directed mutagenesis, we identified an allosteric intradomain network that modulates the Ca2+ affinity depending on the pH, thereby promoting fast ligand release.



INTRODUCTION Many members of the protein family of C-type lectin receptors (CLRs) serve as transmembrane or secreted pattern recognition receptors of the innate immune system and play an important role in regulating the immune response. This pathogen recognition process is based on defined glycan structures present on the surface of the pathogen and, in the case of membrane-bound receptors, recognition can lead to uptake and successive antigen processing and presentation.1 A central Ca2+ ion that is coordinated by a cage of charged amino acid side chains of the carbohydrate recognition domain (CRD) is the mediator of this interaction (Figure 1A) and thereby ultimately the key element in regulating the uptake and routing of the cargo to the endosomal compartment. While the extracellular milieu provides stable Ca2+ concentration and pH, acidification and active Ca2+ export in the early endosome highly reduce the carbohydrate affinity, resulting in cargo release. Consequently, every step from the initial contact, triggering the uptake, to cargo release and sorting, and in some cases receptor recycling, is tightly controlled. The architecture of CLRs reflects this demand for an adaptive ligand binding module with a remarkable evolutionary flexibility to keep up with co-evolving pathogens.2 While the pH dependency for both carbohydrate ligands and Ca2+ binding has been reported for several CLRs, such as DC© XXXX American Chemical Society

SIGN, DC-SIGNR, and the hepatic receptor asialo glycoprotein receptor (ASGPR),3−6 limited structural data are available on the molecular mechanisms underlying the important determinants of cargo uptake and release. Two such determinants have been associated with cargo release upon acidification of the receptor environment. First, some oligomeric CLRs lose their multimerization state, resulting in decreased avidity.7 Second, protonation of the amino acids of the Ca2+ cage releases the ion and abrogates carbohydrate interaction: for the ASGPR, a central histidine residue acts as the pH sensor whose side chain protonation state controls Ca2+ affinity.5,8,9 However, pH dependence of binding is not universal for all CLRs, and even closely related family members show different behaviors, such as DC-SIGN, DC-SIGNR, and LSECtin, wherein only the former two proteins show pH-dependent binding.4 For DCSIGNR, a decrease in pH results in structural or dynamical changes and a partial release of Ca2+,10 but the underlying molecular mechanisms remain elusive. Large conformational changes of the CRD are associated with Ca2+ release in almost all cases for which a structure has been solved in the absence of the Ca2+ in the canonic carbohydrate binding site. A conserved proline, located in the central long loop harboring this site, Received: May 27, 2016

A

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 1. Ca2+ binding of Langerin is highly pH dependent. (A) Cartoon representation of trimeric Langerin ECD (PDB entry: 3KQG34) and expansion of binding site with complexed Ca2+ and dimannoside fragment (PDB entry: 3P5D35). (B) Binding of Langerin ECD to mannan-coated plates decreases rapidly upon acidification, with a pH50 of 5.4 ± 0.1 and a Hill coefficient of 1.7 ± 0.3 (n = 3, SEM as error bars). (C) ITC thermogram of Ca2+ interacting with Langerin CRD and ECD at pH 7, revealing the same affinity for Ca2+, with Kd = 105 ± 15 and 130 ± 50 μM, respectively. (D) ITC thermogram of mannose interacting with Langerin CRD and ECD at pH 7. Again, both proteins have similar affinities, with Kd = 6.2 ± 0.5 and 5.6 ± 1.5 mM, respectively. (E,F) ITC thermogram for Ca2+ binding (E) to Langerin CRD at pH 6 (Kd = 800 ± 150 μM) and (F) to mannose (Kd = 9.2 ± 1.2 mM). All measurements were conducted in triplicate at 298 K, and stoichiometry was fixed to 1.

switches back from cis to trans conformation.8,11−14 For the mannose-binding protein (MBP), this mechanism was proposed to serve as a kinetic trap, preventing rebinding of the cargo and allowing separation of recycling receptors and lysosomal degradation of the payload.12 The trimeric CLR Langerin is a pattern recognition receptor mainly found on Langerhans cells.15,16 It is involved in the detection of a wide set of pathogens including viruses such as HIV,17 measles virus,18 and influenza virus,19 as well as fungi20 and mycobacteria,21 and it recognizes antigens of endogenous origin such as H and B blood group antigens22 and heparin.23,24 Langerin harbors a single Ca2+ site25 embedded in the canonical EPN motif that conveys mannose-type monosaccharide specificity. Langerin is a recycling endocytic receptor that releases its cargo in the early endosome, while the receptor itself accumulates in Rab11+ recycling compartments.15,26 Upon acidification, Langerin remains in trimeric form.27 Langerinmediated antigen uptake is of particular interest, as it has been shown to allow for presentation of exogenous peptide antigens to both CD8+ and CD4+ T-cells.28,29 We therefore set out to elucidate the determinants for Ca2+ binding and release and applied nuclear magnetic resonance (NMR) and isothermal titration calorimetry (ITC) in combination with molecular dynamics (MD) simulation. We demonstrated that Ca2+ binding is pH dependent, whereas the effect of acidification on carbohydrate recognition is limited at access Ca2+ concentrations. We attribute the fine-tuning of the Ca2+ affinity to a robust and pH-sensitive allosteric network that couples the motion of the binding loop with that of an adjacent short loop. We identified H294 as a partial pH sensor. This histidine, however, does not coordinate Ca2+ directly but

instead is a central hub of the allosteric network which couples the two loops. The proposed mechanism is distinctively different from the two known mechanisms of Ca2+ affinity control in that it does not rely on a single (switch-like) structural change, but the flexibility of the entire binding loop regulates Ca2+ affinity, which in turn is regulated by an extended allosteric network. Thus, the receptor architecture not only serves as a scaffold for the carbohydrate binding site but takes an active role in cargo uptake and release.



RESULTS

Carbohydrate Binding to Langerin Is Ca2+-Dependent, while Ca2+ Affinity Itself Is pH-Dependent. To investigate the structural determinants of cargo binding and release, we first analyzed the pH dependence of carbohydrate binding to human Langerin extracellular domain (ECD). The affinity to mannan, a natural fungal polysaccharide ligand, dropped rapidly with decreasing pH, with a half-maximum at pH 5.4 ± 0.1, and was abolished below pH 5, in accordance with previous reports (Figure 1B).27 To further dissect whether this loss of binding was due to decreased affinity for the carbohydrate ligand itself or for the central Ca2+ coordinating the glycan, we determined the dissociation constants for mannose as a simple monovalent ligand and the essential cofactor Ca2+ at both pH 7 and pH 6 by ITC (Figure 1C−F) for the CRD. Additional 1H−15N HSQC NMR titrations confirmed these results (Figure S1). The affinity for mannose decreased slightly from 6.2 ± 0.5 to 9.2 ± 1.2 mM at saturating Ca2+ concentrations at pH 6 and 7, respectively. However, the affinity for Ca2+ dropped significantly from 120 ± 65 to 800 ± 150 μM (Figure 1C,E). Moreover, at pH 7, CRD and B

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society ECD share almost identical affinities for both Ca2+ and mannose (Figure 1D). Thus, it can be assumed that there is no major interdomain cooperative behavior involved in the recognition of simple monovalent ligands. pH Dependence of Ca2+ Affinity Is Likely under Intradomain Allosteric Control. Surprisingly, during NMR titration experiments, we observed pronounced changes in chemical shifts of many resonances at both pH 6 and 7, which was unexpected for a rigid protein with a single binding site (Figure 2A). To gain structural insight, 92% of the backbone resonances of the holo form of Langerin CRD at pH 6 were assigned (BRMB entry: 26791, Figure S2). The long loop and the β4 strand remained unassigned, presumably due to unfavorable dynamics of this region, as observed for other CLRs.30,31 We identified 43 residues at pH 6, and 48 at pH 7,

outside a 5 Å radius of the binding site that were affected by Ca2+ binding (Figure 2C). As these two sets were 95% identical, we concluded that the same residues are involved in Ca2+ binding at both pH values. These perturbations are located predominantly in proximity of the glycan binding site, with the most prominent effects in the β2 and β2′ strand and the enclosed short loop (residues 255−265, Figure 2C). Intriguingly, these amino acids were pH responsive in the holo form (Figure S3A). Closer inspection of the chemical shift changes revealed that the short loop in the holo form has distinct conformations at pH 6 and pH 7 (Figure S3B−D). Analysis of 28 mammalian Langerin homologues revealed a significantly higher evolutionary conservation of those residues compared to the rest of the sequence (Figures 2C and S4), indicative of a conserved function. Moreover, we observed that many peaks undergo severe line broadening during Ca2+ titration at both pH values, which is indicative for intermediate fast exchange on the NMR time scale that might be caused by slow kinetics of binding and allosteric structural rearrangements on the micro- to millisecond time scale32 (Figure S5A). This observation might explain the deviations between the dissociation constants determined from ITC and NMR (Figure S5B).33 Taking these findings together, we determined that pH-dependent ligand release is dominated by decreased Ca2+ affinity. In turn, Ca2+ binding affects a large network of residues in a pHdependent manner, indicating an allosteric control mechanism. Apo Langerin CRD Undergoes pH-Independent Prolyl Cis/Trans Isomerization. CLR Ca2+ coordination is governed by a cis/trans isomerization of a conserved proline of the EPN motif in the long loop forming the Ca2+ binding site.8,11−14 To exclude that changes in cis/trans isomerization account for the observed differences in affinity at both pH values, we analyzed the isomerization state of the long loop (residues 283−295). The 1H−15N HSQC spectrum of the Ca2+ unbound form (apo) showed additional peaks which we could attribute to the trans state of P286 in slow exchange with the help of a P286A mutant. We assigned 22 resonances of the trans state (Figure S6A). Notably, many of the resonances corresponding to the cis state of these interchanging residues experience significant chemical shift changes upon addition of Ca2+ (Figure 2B), substantiating our previous observation of an extended network of communicating residues. Surprisingly, the average ratio of the peak intensities of the cis and trans forms showed that 75 ± 10% of the apo form are in cis prolyl bond conformation (Figure S6B), which is in contrast to other CLRs, where the majority of the apo population is found to be trans.8,11−14 Moreover, the cis/trans ratio and the chemical shift differences between the two states were unaffected by pH (Figure S6C) and do not explain the affinity differences upon acidification. Based on the 15N chemical shift differences, we estimated an upper limit for the exchange rate of 1 s−1 (Table S1).12 Furthermore, upon addition of Ca2+, only the resonances of the cis form shifted, while the volume of the peaks of the trans form decreased. Hence, we conclude that only the cis form binds Ca2+. These data are in line with previous reports on CLRs only with the prolyl bond in cis conformation being Ca2+ binding competent.36,37 Additionally, these data suggest that the apo trans form is in slow exchange with the holo cis form, likely via the apo cis form. Thus, the decrease in trans population is caused by a shift of the equilibrium toward the holo cis form under saturating Ca2+ concentrations. These

Figure 2. Ca2+ binding affects several residues remote from the binding site. (A) Overlay of 1H−15N HSQC NMR spectra of apo and holo Langerin CRD at pH 6 and pH 7 at 298 K (holo forms are labeled in blue shades and apo forms in red shades; pH 6 forms are labeled in the stronger color hue). (B) Spectra expansions of K257 and E285 backbone resonances at increasing Ca2+ concentrations from 0 to 10 mM at pH 6 in the intermediate exchange regime. (C) Top: chemical shift perturbations induced by addition of Ca2+ at pH 6. Assigned residues and resonances in slow exchange in the apo form are highlighted gray and green, respectively. Bottom: evolutionary conservation of allosteric network residues (labeled purple) among 26 mammalian species is significantly higher than that of the residues that do not participate in the network (gray) (p < 0.005, paired Wilcoxon signed ranks test). (D) Amino acids affected by Ca2+ addition (chemical shift perturbation (CSP) > 0.02 ppm) highlighted on the Langerin structure (PDB entry: 3P5F; color-coded according to the CSP; unassigned residues are colored salmon). C

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 3. Conformational dynamics of the Langerin CRD is restricted to the long and the short loops. (A) Cartoon representation of the trans apo (blue) and cis apo (green) conformations superimposed onto a Langerin crystal structure (gray, PDB entry: 3P5H35), highlighting the altered conformational space in trans apo form (I) and the disordered α3-helix (II). (B) Root-mean-square fluctuations (RMSFs) of the backbone nitrogen atoms averaged over the simulation (2 μs) for the three Langerin CRD forms. The largest RMSF values are observed for residues of the short loop (residues 255−265) in all forms, while residues of the Ca2+ binding site undergo higher structural fluctuations only in the trans apo form. (C) The chemical environment of A258 is perturbed by changes in backbone flexibility of the adjacent amino acid D263. In the apo form, the short loop adopts a more opened conformation (green) compared to the holo crystal structure (gray). Expansions I and II highlight the preferred conformations of A258 and D263 in the holo and apo forms, respectively. (D) Autocorrelation functions calculated for the distance between Cαatoms of M260 and G290. A double-exponential fit (dashed line) shows concerted mobility of the short and the long loops on the nanosecond time scale. (E) 15N backbone relaxation rate constants R1 and R2 and hetero NOE of apo (red, upper panel) and holo (blue, lower panel) Langerin CRD forms (600 MHz, pH 6, 299 K). The protein exhibits uniform relaxation rate constants, except in the short loop and the long loop.

drastically in these two conformations (Figure 3C). Thus, the observed chemical shift perturbation upon Ca2+ binding is due to the removal of the minor conformation from the ensemble. We suspected that the dynamics of the short loop could be correlated to the mobility of the proximal long loop for two reasons: (i) the long loop directly interacts with the Ca2+ ion, but the largest chemical shifts upon Ca2+ binding occur in the short loop, and (ii) if Ca2+ is indeed under allosteric control, then other flexible regions of the protein are likely involved. We tested this by computing autocorrelation functions of the distance between the two loops for several residue pairs. After a fast initial decay, the distance autocorrelation function decayed approximately exponentially, with time scales of 5−6 ns. This was found for both the cis apo and the cis holo forms (Figure 3D). This coupling on the nanosecond time scale is indicative of structural transitions that require a cooperative movement of the two loops. While this does not prove allostery in itself, it strongly hints at a pathway by which information (e.g., presence or absence Ca2+) can be transmitted. To gain more experimental insight into Langerin dynamics, NMR 15N-backbone relaxation data were acquired at two field strength for both holo and apo forms (Figures 3E and S8A). Regardless of the presence of Ca2+, we observed a pronounced reduction in the R2 relaxation rate and the hetero NOE in the short loop compared to the remainder of the protein. Model-free analysis of the relaxation data was unsuccessful, likely due to major motions on slower time scales. However, the hetero NOE is sensitive to changes in local correlation times and can thus serve as a good estimate for internal motions. This reduction is therefore indicative for an increased mobility on the pico- to nanosecond time scale, which is in agreement with the conformational flexibility of the short loop in our simulations. In the apo form, an overall increase in R2 rate constants compared to the holo form (16 ± 3 vs 15 ± 3 s−1

results, taken together, show that the cis prolyl conformational state of Langerin is in slow exchange and is highly preformed independent of pH, and thus do not explain the pH-dependent affinity drop. Langerin Shows Increased Nanosecond Dynamics in the Short and Long Loops. In the absence of an X-ray crystal structure of the apo form and because the long loop was unresolved by NMR, we turned to MD simulations to gain more insight into how such large chemical shift perturbations could propagate from the Ca2+ binding site. We performed microsecond all-atom simulations in explicit water for all structures in the Ca2+-binding equilibrium at pH 7: cis apo, cis holo, and trans apo Langerin. Excellent agreement between MD simulations and NMR data was demonstrated by comparing chemical shifts and 3JHNHA coupling constants for both cis apo and holo forms (Figure S7). The overall structure of the cis apo form is highly similar to the structure of the cis holo form. Both simulations showed a rigid protein backbone where only the short loop exhibited increased backbone dynamics (Figure 3A,B). Additionally, the α3-helix became destabilized in the cis apo form. Although the trans apo form also showed a largely rigid backbone, the long loop sampled a distinctively different conformational space than in the two cis forms (Figure 3C). The latter simulation also showed that Ca2+ binding is sterically prohibited in the trans form, confirming the NMR analysis. The large chemical shifts upon Ca2+ binding of residues remote from the Ca2+ binding site (found both in MD and in NMR) could be attributed to mostly minor local structural changes. For example, in the crystal structure, the backbone torsion angles of D263 are in the β-sheet conformation (shown in gray in Figure 3C). This conformation is sampled in both apo and holo forms. However, in the apo form, additionally the α-helix region is populated. The chemical environment of the N−H bond of A258 differs D

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 4. Mutual dependence graphs revealed correlated movement of the long and the short loops. Network representation of the mutual dependence graphs computed for the trans apo (left), cis apo (middle), and cis holo forms (right). Edge thickness corresponds to the aggregate NMI values (threshold 0.02). The edge between the hub residues H294 and K257 is highlighted (green). Nodes are colored according to CSP values observed upon Ca2+ binding. Nodes representing the long and short loops are marked by red and blue borders, respectively. The aggregated NMI values for H294−K257 decrease from 0.08 to 0.05 to 0.02 from left to right.

Figure 5. Single-residue mutations in Langerin CRD induce pronounced chemical shift changes remote from the mutation site. (A) Cartoon representation of Langerin CRD (PDB entry: 3P5F), with the mutated residues shown as red spheres and Ca2+ ion in black. (B) CSP studies of mutant apo and holo forms (left and right panel) compared to the WT, showing a distinct CSP pattern. Bottom row shows the WT binding to Ca2+. CSPs of the mutants upon addition Ca2+ (right panel) show a high conservation of the WT pattern except in K257A and H294A. Assigned residues are highlighted gray. All data were recorded at pH 6, 298 K. (C) Cartoon representation of the Ca2+ interaction network, with the mutated residues shown as red spheres and the Ca2+ ion in black.

at 600 MHz) as well as a field dependence of these rate constants of about 1 s−1 in both forms (Table S2) indicated the existence of dynamics on the micro- to millisecond time scale.38 Taken together, our MD simulation and experimental data are in good agreement and reveal that holo and apo Langerin CRD forms are structurally very similar and rather rigid, with the exception of the short and long loop regions. In both forms, the short loop moves correlated with the long loop on the nanosecond time scale, indicating a coupling of the loop dynamics.

An Allosteric Network Connects Distal Residues in Langerin and Decreases Ca2+ Affinity. Our observations strongly indicate an allosteric mechanism but do not directly prove its existence. Structural allostery exists if the conformations sampled by a region A depend on the current conformation in another region B of the protein. The mutual information directly measures to which degree the conformations in A are determined by the current conformation in B. It can directly be interpreted as the amount of information which is transmitted between the regions and has been used previously to detect allosteric networks in MD simulations of E

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society Table 1. Binding Affinities of Langerin CRD to Ca2+ at pH 6 and pH 7, Obtained by ITC and NMR Measurements ITCc

NMR

mutant

Kd (μM)

ΔG (kJ mol−1)a

WT (pH 6) WT (pH 7) Q239A (pH 6) K257A (pH 6) D263A (pH 6) S265A (pH 6) H294A (pH 6) H294A (pH 7) N297A K299A

800 ± 150b 105 ± 15b

−17.8 ± 34 −22.9 ± 3.3

200 ± 20

−21.2 ± 0.3

125 ± 5 35 ± 15

−22.4 ± 0.9 −25.5 ± 8.5

Kd (μM) 620 160 640 265 620 570 110

± ± ± ± ± ± ±

50b 30 40 30 30 30 10

770 ± 30 210 ± 20

ΔG (kJ mol−1)a −18.4 −21.8 −18.3 −20.6 −18.4 −18.6 −22.7

± ± ± ± ± ± ±

1.0 3.4 1.0 1.7 1.1 0.8 0.8

−17.9 ± 0.5 −21.1 ± 1.3

Free energies were calculated according to the relation ΔG = −RT ln Kd, with R = 8.1314 kJ mol−1 K−1 and T = 300 K. bError given as standard deviation of three independent measurements. cStoichiometry was set to 1 to determine Kd using a one-set-of-sites binding model. a

proteins.39,40 We computed the normalized mutual information (NMI) between the φ−ψ torsion angle space (i.e., Ramachandran planes), and between the side chain torsion angles of all pairs of residues. The results of the various NMIs per residue pair were added to obtain an aggregated measure for the interresidue correlation, revealing a common extended network of correlated residues within the three structures of Langerin (cis apo, cis holo, and trans apo) (Figures 4 and S9). Furthermore, we found that the physical interactions underlying these correlated motions are propagated by a highly stable hydrogen bond network (Figure S10, Table S1). We mapped the residues that experienced chemical shift perturbations during Ca2+ titration (cf. Figure 2B) to the same mutual information graph. Our findings emphasize a high similarity in connectivity but deviate in the magnitude of the underlying correlations. These decrease when the system is shifted from trans to cis apo form, and further decrease in the fully Ca2+-bound holo form, suggesting an overall loss in correlation upon Ca2+ binding. If such an allosteric network existed, mutations should introduce perturbations that propagate through the network and result in chemical shift changes.41 To probe the interaction between the long and short loops and their connection to other regions of the fold, 10 mutants were prepared that were folded and bound Ca2+, as was observed by NMR (Figures 5A and S11). We quantified the chemical shift perturbations induced by mutations in their apo form with respect to the wild type (WT). For the majority of the mutants, perturbations were not restricted to the neighborhood of the mutation site but also affected remote residues. The perturbed residues typically were members of the network of correlated residues obtained from MD analysis (Figure 5B, left). Comparison of the holo forms of the mutant with the holo WT yielded the same picture for most mutants, indicating that the network persists in the Ca2+ bound state (Figure 5B, middle). Finally, Ca2+ titration perturbed similar residues in most mutants compared to the WT lectin (Figure 5B, right), allowing for the definition of a robust core network of 24 amino acids (Figure 5C). Notably, H294A and K257A mutants binding to Ca2+ yielded a distinct pattern of chemical shift perturbations compared to the WT, with an about 2-fold reduction in the number of affected residues, indicating that the allosteric network is corrupted in these mutants. To assess whether the mutations also affected Ca2+ affinity, we determined the dissociation constants for seven mutants by NMR titrations at pH 6 (Table 1, Figure S11). Intriguingly,

three mutants (H294A, K257A, K299A) showed increased affinity at pH 6, wherein H294A experienced the strongest effect (Kd = 110 μM). Since none of the mutants had an altered cis/trans isomerization of P286 at pH 6, we could exclude a contribution of the allosteric network to a conformational stabilization of either isomer at this pH (Table S3). Replacing K299 by alanine removes a positive charge in the vicinity of the Ca2+ binding site, which explains the higher Ca2+ affinity of the K299A mutant. The high Ca2+ affinity in the two other mutants is likely linked to the observed loss of allosteric communication. Hence, our data suggest that the allosteric network in Langerin downregulates Ca2+ affinity. H294 Is a Key Player in the Allosteric Network and Couples the Long and Short Loops. Residue H294 is a key component of the allosteric network because it is a central node in all three mutual information graphs connecting the clusters of the short and long loops. Notably, mutating this amino acid affected many residues that were also affected by Ca2+ binding in the WT (Figure 5A, left). To investigate the structural origin of these perturbations, we performed a chemical shift projection analysis (CHESPA).42 Fractional shifts of the equilibrium population toward the holo or apo form can be delineated from this analysis on a per-residue basis by the magnitude and angle of the chemical shift change. The CHESPA analysis revealed a set of positive shifts along the WT apo-holo vector for the short loop region, whereas the residues flanking the long loop and the Ca2+ binding site shifted in the opposite direction, indicating an overall decoupling of the two loops (Figure 6A). For the other mutants, however, no clear pattern could be derived, suggesting a more complex mechanism (Figure S13). To further investigate the decoupling of the loops, we performed 2 μs MD simulations of the apo and holo H294A mutants. The distance distributions between the two loops and the autocorrelation function of the distance time series were calculated (Figure 6B,C). The distributions of the four systems (apo and holo WT, apo and holo H294A) show three major maxima, which we denotedgoing from smaller to larger distancesclosed, intermediate, and open states. In the closed state, the two loops are in direct van der Waals contact. However, the relative populations of these maxima, in particular of the closed state, vary strongly between the four structures. In the holo H294A mutant, the closed state is not populated at all, and the autocorrelation function decays much faster than in the other three systems. Thus, the ability to form the closed state seems to be essential for the coupling of the two loops. F

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 6. H294A mutant of Langerin CRD alters dynamics of the apo form and decouples the long and short loops. (A) CHESPA analysis of H294A apo at pH 6. Most changes induced by the mutation align very well with the reference vector (|cos(θ)| ≈ 1, upper panel). Lower panel shows the projection of the vector lengths. (B) The short loop can adopt three distinct conformational states as monitored by the distance between the Cαatoms of M260 and G290: closed (C), intermediate (I), and opened (O). (C) Autocorrelation functions (ACFs) computed for the same distance as in (B) (dashed line shows the double-exponential fit). The ACF of the holo H294A mutant showed the fastest decay (3.9 ns), suggesting a decoupling of the correlated movement of the two loops. (D) Overlay of 1H−15N HSQC NMR spectra of the apo forms of WT (grey) and H294A (red) CRD at pH 6, 298 K. The mutant spectrum resolves additional peaks, likely being the previously unassigned residues of the long loop. (E) Mutual dependence graphs computed for residues 255−310 of H294A cis apo (left panel), cis holo (center panel), and cis apo WT with protonated H294 side chain (right panel). The hub residues H294 and K257 are highlighted in green. Mutation of H294 led to the loss of correlated movement along H294−K257 axis. (F) Two hydrogen bonds connect the hub residues K257 and H294, coupling the movement of the two loops. (G) NMR chemical shift perturbations of WT and H294A at pH 7. While the mutant and WT apo forms still differ, the holo forms have identical structures.

movement of the two loops, further evidencing that the decoupling of the loops is connected with the increased affinity for Ca2+ (Figure 6E). Taken together, MD and NMR showed partial decoupling of the loop motions upon H294 mutation, which results in increased mobility of the long loop and increased affinity toward Ca2+. H294 Partially Entails pH Responsiveness. The prominent role of H294 in the allosteric network, its elevated affinity for Ca2+ at pH 6, and the strong pH dependency of Ca2+ binding in the WT led us to hypothesize that H294 is a pH sensor similar to H256 in rat ASPGR.5 First, we calculated pKa values for the apo and holo forms using two orthogonal

Additionally, we noticed during our NMR titration experiments of the H294A mutant increased peak intensities and the appearance of several additional peaks (Figure 6D), which likely belong to unassigned residues in the long loop. Hence, the removal of the H294 side chain shifted the exchange regime of the long loop on the NMR time scale, further substantiating that the dynamics of this region changed notably. Although we observed decreased chemical shift perturbations in the short loop region upon Ca2+ addition in the H294A mutant, the communication was not completely abrogated, supporting the notion that the network is robust.41 Mutual dependence graphs computed for the H294A mutant revealed a loss of correlated G

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society algorithms making use of representative MD snapshots (Table S4), awarding pKa values close to 6 for both forms, implying that approximately half the population is protonated at pH 6. To assess whether protonation of H294 was solely responsible for the pH dependence of Ca2+ affinity, we determined the dissociation constants of the H294A mutant by ITC at both pH 6 (Kd = 125 ± 5 μM) and pH 7 (Kd = 36 ± 15 μM) (Figure S14A). Moreover, comparison of the chemical shifts of both holo and apo forms at pH 6 and 7 shows that the pH responsiveness in the short loop in the holo form is conserved in the absence of the H294 side chain (Figure S14B). Both findings showed that the H294 side chain is not the sole pH sensor. However, comparing the pH-dependent affinity increase of both WT and mutant revealed that the WT is more pH responsive than the mutant, implying a partial contribution of H294 to the pH effect. We used MD simulation of the apo protein with protonated H294 (apo cis H294+) to gain more insight into the pH effect. The side chain of H294 forms a hydrogen bond with the side chain of K257 in the short loop, which was populated about 30−35% in the three WT simulations (apo trans WT, apo cis WT, holo cis WT). It is backed up by a second hydrogen bond from the backbone amide hydrogen of K257 to the carboxyl oxygen of H294, which is populated 51−56% in the three WT simulations (Figure 6F, Table S1). Our MD simulations of apo cis H294+ showed that the side chain hydrogen bond is indeed disrupted but the backbone hydrogen bond was maintained. Moreover, since both side chains of K257 and H294 are protonated in apo cis H294+, the two loops experience a Coulomb repulsion and are pushed apart (Figure 6E), which can be observed in the drastically altered distance distribution of the two loops for apo cis H294+ (Figure 6B). In particular, the closed state is not visited at all, and the two loops do not seem to be in direct van der Waals contact. Correspondingly, the autocorrelation function differs from those of the WT and H294A mutant structures (Figure 6C). The initial decay is more pronounced, after which the autocorrelation function decays relatively slowly, with a time scale of 5.8 ns. Thus, protonation of H294 strongly changes the conformations and the dynamics of the two loops. Notably, the effect of protonation of H294 was much stronger than that of the H294A mutation. Thus, we conclude that the contribution of H294 to the pH sensitivity of Langerin is predominantly caused by the presence of a second positive charge between the loops and not by the disruption of the hydrogen bond. Finally, we noted that the WT and H294A holo chemical shifts almost completely overlapped at pH 7 (Figure 6G), which was not the case at pH 6 (Figure 4, middle). This is a strong indicator for structural similarity of the WT and the H294A mutant at pH 7 and a further evidence for a pHsensitive component encoded in the allosteric network independent from H294. The Allosteric Network Destabilizes the Holo State with Decreasing pH. To dissect the contributions of the single components to the pH-dependent regulation of Ca2+ binding by the allosteric network, we constructed a thermodynamic model based on the free energy differences of Ca2+ binding in the WT and in the mutants at both pH 6 and 7 (Table 1, Figures 7 and S14A,C). We calculated the free energy differences, which ranged from −17.8 kJ mol−1 in WT at pH 6 to −25.5 kJ mol−1 in H294A at pH 7. Our analysis showed that the pH sensitivity of the Ca2+ affinity in WT Langerin (in total ΔΔG = 5.1 kJ mol−1) is a combination of

Figure 7. Contributions to the Ca2+ binding free energies in WT, H294A, K257A, K299A, and Q239A.

two pH sensors: H294 and an unknown sensor. The pH sensitivity of the H294A mutant is hence only due to the second pH sensor and amounts to ΔΔG = 3.1 kJ mol−1. Consequently, the contribution of H294 to the pH sensitivity in WT Langerin is 2 kJ mol−1. The overall Ca2+ affinity in the H294A mutant is increased compared to that of WT Langerin. We attributed this increase to the disruption of the allosteric network and estimate its magnitude to be ΔΔG = 4.6 kJ mol−1. For the K257A mutant, this effect amounts to ΔΔG = 3.4 kJ mol−1, which leaves an unexplained difference compared to the H294A mutant of ΔΔG = 1.2 kJ mol−1. Allosteric communication is based on a multitude of physical interactions: presence and relative population of hydrogen bonds, presence and positions of charges, van der Waals contacts, loop flexibility, etc. To dissect the individual contributions of these interactions to the allosteric network requires the detailed investigation of a series of mutants at various pH values. However, we can speculate that the second pH sensor is a protonation site in the vicinity of the Ca2+ binding site. In the K299A mutant at pH 6, this site is protonated, but by removing the charged lysine at position 299, the effect of this second pH sensor is muted. Thus, compared to WT Langerin at pH 6, the Ca2+ affinity of K299A at pH 6 should be increased (ΔΔG = 3.1 kJ mol−1). This is in very good agreement with the measured binding constant. But again, detailed analysis of other mutants and pH values is needed to test this hypothesis. Finally, Q239 is not part of the allosteric network and is far away from the binding site. The mutant Q239A serves as a control and should behave as the WT Langerin. Indeed, the Ca2+ binding constants differ by only 0.5 kJ mol−1, which is within the experimental accuracy. This also confirms that the increase in binding constants in H294A and K257A is in fact due to the disruption of the allosteric network and not an artifact of the mutation itself.



DISCUSSION Langerin is an endocytic receptor that serves as a pattern recognition receptor, initiating the immune response to invading pathogens. Langerin is a good model to further our insight into the underlying molecular mechanisms of cargo binding and release. Our results reveal the absence of interdomain allosteric communication between the CRDs for interactions with monovalent ligands, as previously observed for DC-SIGN and DC-SIGNR,43 allowing for the isolated study of the CRD using NMR spectroscopy. Hence, we were able to H

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society identify Ca2+ affinity and its pH sensitivity as the two major contributors of ligand release. Under extracellular conditions, pH 7, and 1−2 mM Ca2+ concentration, Langerin has a high affinity for Ca2+, shifting the binding equilibrium almost completely to the holo state, an observation in line with other CLRs.8 Upon acidification to pH 6, resembling the milieu of the early endosomal compartment, the affinity for Ca2+ decreases significantly, thus favoring ligand release. For the mannose-binding protein, a kinetic trap mechanism based on the very slow prolyl bond isomerization of the central proline residue was suggested that prevents rebinding of the cargo to a recycling receptor.12 Remarkably, a unique feature of Langerin is its high (75%) pH-independent cis prolyl bond conformation in the absence of Ca2+. This degree of pre-organization of the long loop harboring the central proline is high compared to other CLRs, where only 10% was reported for Tetranectin.8,11−14 Hence, kinetically trapping the receptor in a binding-incompetent state is seemingly not at play for Langerin, and rebinding to the recycling receptor is likely prevented by decreased Ca2+ affinity and potentially active Ca2+ export in the early endosome.44 Surprisingly, our analysis provided evidence from complementary methods for the existence of an extended allosteric network of communication amino acid side chains. NMR chemical shift perturbations of backbone resonances of Langerin by Ca2+ titration are located in sites distal from the actual Ca2+ recognition site, suggesting a propagation of conformational changes in Langerin upon cofactor binding. Conversely, perturbing these remote sites using single-point mutations caused these changes to propagate along the same routes through the protein, similar to other allosteric proteins.41 None of the single-point mutations was able to completely abrogate the network, suggesting that it is robust.41 Moreover, our mutant analysis then unraveled a core network of amino acids which is evolutionarily conserved in Langerin. These data are substantiated by an extensive analysis of several microsecond MD simulations using mutual information theory. Taken together, NMR, MD simulations, and evolutionary considerations led to the identification of an allosteric network in Langerin being associated with Ca2+ binding. Among the mutations, H294A turned out to be particularly informative, revealing that the network couples the mobility of the short and the long loops, thereby weakening Ca2+ affinity, possibly by slowing down the motion of the long loop. Strikingly, the magnitude of correlated mobility observed per residue decreased from the trans to the cis apo state, being even further decreased in the holo form (Figure 4). These data suggest that the allosteric communication is impaired upon Ca2+ binding. Upon mutation of either H294 or its direct interaction partner K257, loop coupling is decreased and Ca2+ affinity is increased. We propose that the coupling of the loops restrains the long loop close to the short loop, and upon decoupling, the long loop has improved Ca2+ coordination (Figure S16). Notably, K257 and H294 side chains highly overlap with the Ca2+-1 and -3 sites of DC-SIGN, a related endocytic CLR (Figure S17). These data hint for an evolutionary mechanisms to substitute for the charge effects introduced by the allosteric Ca2+ ions in other CLRs that finetune affinity for the primary Ca2+. Hence, the allosteric network fulfills now this function of fine-tuning the affinity, a hypothesis currently under investigation in our laboratories. We further attempted to identify the pH-dependent mechanism for Ca2+ release. As H294 was identified as a

central player of the network, we asked whether its protonation state might serve as a regulator in analogy to the ASGPR, a hepatic CLR for which H256 was previously reported to play such role.5 Our MD simulation suggested that H294 protonation at least partially decouples the loop dynamics (Figure 6B). To further delineate the pH-dependent and -independent contributions to Ca2+ affinity, we constructed a thermodynamic model based on the affinities of the WT and mutant data at both pH 6 and 7 (Figure 7, Table 1). Strikingly, H294 only partially contributes to the pH dependence of the Ca2+ affinity, as could be seen from the persisting pH dependence of the H294A mutant. Other potential pH sensors are the members of the Ca2+ cage E285, E293, and D308, for which remarkably high pKa values were calculated (Figure S18, Table S4). In particular, the former two residues show similarity to an aspartyl dyad with elevated pKa values, similar to residues found in BACE-1 and HIV-1 protease.45,46 Several physical interactions change upon protonation of H294: (1) the side chain hydrogen bond to K257 is broken, (2) the dynamics of the long loop and the short loop is decoupled, and (3) a second positive charge is introduced between the two loops, leading to a strong Coulomb repulsion between K257 and H294. Not all of these changes necessarily decrease the Ca2+ affinity, but the overall change in Ca2+ affinity seems to be in subtle balance between opposing effects. It is difficult to probe these effects individually, since mutating one of the interaction partners, K257 or H294, also affects the reference state at pH 7. In particular, the loops are at least partially decoupled in the two mutants. Moreover, the direct van der Waals contact between the side chains of K257 and H294, which stabilizes the closed conformation of the loops in the WT at pH 7, is perturbed upon mutation, thereby introducing a fourth effect. Finally, K257A and H294A differ in their charge states, as K257 is protonated at pH 7 and pH 6, whereas H294 can only be protonated at pH 6. This likely explains the observed differences in Ca2+ affinity in the two mutants. In addition to these insights into the role of the allosteric network, we were also able to dismiss potential functions. First of all, we did not observe any marked difference of the cis/trans prolyl bond isomerization of the CRD at pH 6 for any mutant (Table S3), rendering this isomerization a network-independent conformational rearrangement. Moreover, the network is restricted to residues close to the ligand binding site, indicating that it is not involved in transmitting ligand binding information through the protein into the cell, initializing cellular signaling pathways.37 Finally, as several carbohydrate binding proteins have been suggested to compensate loss of conformational entropy upon ligand binding by increasing conformational entropy at remote sites of the protein, we investigated such a hypothesis using our models.47−49 However, without additional data, which is beyond the scope of this work, no clear conclusions could be drawn. Overall, Langerin experiences remarkable dynamics on various time scales, as revealed by NMR and MD simulations. The receptor is rather rigid under extracellular conditions, and its mobility is restricted to the coupled movement of the long and short loops, likely on the nanosecond time scale, as inferred from NMR measurements. Upon endocytosis and hence Ca2+ release, additional micro- to millisecond dynamics appear, inferred from our NMR relaxation experiments (Figures 3E and S6). Similar observations have been reported for other CLRs.10,13 The mobility of Langerin on the pico- to nanosecond time scale did not change under these conditions, I

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

(9) Feinberg, H.; Torgersen, D.; Drickamer, K.; Weis, W. I. J. Biol. Chem. 2000, 275, 35176. (10) Probert, F.; Mitchell, D. A.; Dixon, A. M. FEBS J. 2014, 281, 3739. (11) Ng, K. K.; Park-Snyder, S.; Weis, W. I. Biochemistry 1998, 37, 17965. (12) Ng, K. K.; Weis, W. I. Biochemistry 1998, 37, 17977. (13) Poget, S. F.; Freund, S. M.; Howard, M. J.; Bycroft, M. Biochemistry 2001, 40, 10966. (14) Nielbo, S.; Thomsen, J. K.; Graversen, J. H.; Jensen, P. H.; Etzerodt, M.; Poulsen, F. M.; Thogersen, H. C. Biochemistry 2004, 43, 8636. (15) Valladeau, J.; Ravel, O.; Dezutter-Dambuyant, C.; Moore, K.; Kleijmeer, M.; Liu, Y.; Duvert-Frances, V.; Vincent, C.; Schmitt, D.; Davoust, J.; Caux, C.; Lebecque, S.; Saeland, S. Immunity 2000, 12, 71. (16) Romani, N.; Clausen, B. E.; Stoitzner, P. Immunol. Rev. 2010, 234, 120. (17) de Witte, L.; Nabatov, A.; Pion, M.; Fluitsma, D.; de Jong, M. A.; de Gruijl, T.; Piguet, V.; van Kooyk, Y.; Geijtenbeek, T. B. Nat. Med. 2007, 13, 367. (18) van der Vlist, M.; de Witte, L.; de Vries, R. D.; Litjens, M.; de Jong, M. A.; Fluitsma, D.; de Swart, R. L.; Geijtenbeek, T. B. Eur. J. Immunol. 2011, 41, 2619. (19) Ng, W. C.; Londrigan, S. L.; Nasr, N.; Cunningham, A. L.; Turville, S.; Brooks, A. G.; Reading, P. C. J. Virol. 2016, 90, 206. (20) de Jong, M. A.; Vriend, L. E.; Theelen, B.; Taylor, M. E.; Fluitsma, D.; Boekhout, T.; Geijtenbeek, T. B. Mol. Immunol. 2010, 47, 1216. (21) Hunger, R. E.; Sieling, P. A.; Ochoa, M. T.; Sugaya, M.; Burdick, A. E.; Rea, T. H.; Brennan, P. J.; Belisle, J. T.; Blauvelt, A.; Porcelli, S. A.; Modlin, R. L. J. Clin. Invest. 2004, 113, 701. (22) Hsu, T. L.; Cheng, S. C.; Yang, W. B.; Chin, S. W.; Chen, B. H.; Huang, M. T.; Hsieh, S. L.; Wong, C. H. J. Biol. Chem. 2009, 284, 34479. (23) Chabrol, E.; Nurisso, A.; Daina, A.; Vassal-Stermann, E.; Thepaut, M.; Girard, E.; Vives, R. R.; Fieschi, F. PLoS One 2012, 7, e50722. (24) Munoz-Garcia, J. C.; Chabrol, E.; Vives, R. R.; Thomas, A.; de Paz, J. L.; Rojo, J.; Imberty, A.; Fieschi, F.; Nieto, P. M.; Angulo, J. J. Am. Chem. Soc. 2015, 137, 4100. (25) Thepaut, M.; Valladeau, J.; Nurisso, A.; Kahn, R.; Arnou, B.; Vives, C.; Saeland, S.; Ebel, C.; Monnier, C.; Dezutter-Dambuyant, C.; Imberty, A.; Fieschi, F. Biochemistry 2009, 48, 2684. (26) Mc Dermott, R.; Ziylan, U.; Spehner, D.; Bausinger, H.; Lipsker, D.; Mommaas, M.; Cazenave, J. P.; Raposo, G.; Goud, B.; de la Salle, H.; Salamero, J.; Hanau, D. Mol. Biol. Cell 2002, 13, 317. (27) Stambach, N. S.; Taylor, M. E. Glycobiology 2003, 13, 401. (28) Idoyaga, J.; Cheong, C.; Suda, K.; Suda, N.; Kim, J. Y.; Lee, H.; Park, C. G.; Steinman, R. M. J. Immunol. 2008, 180, 3647. (29) van der Vlist, M.; Geijtenbeek, T. B. Immunol. Cell Biol. 2010, 88, 410. (30) Probert, F.; Mitchell, D. A.; Dixon, A. M. FEBS J. 2014, 281, 3739. (31) Pederson, K.; Mitchell, D. A.; Prestegard, J. H. Biochemistry 2014, 53, 5700. (32) Williamson, M. P. Prog. Nucl. Magn. Reson. Spectrosc. 2013, 73, 1. (33) Feeney, J.; Batchelor, J. G.; Albrand, J. P.; Roberts, G. C. K. J. Magn. Reson. 1979, 33, 519. (34) Feinberg, H.; Powlesland, A. S.; Taylor, M. E.; Weis, W. I. J. Biol. Chem. 2010, 285, 13285. (35) Feinberg, H.; Taylor, M. E.; Razi, N.; McBride, R.; Knirel, Y. A.; Graham, S. A.; Drickamer, K.; Weis, W. I. J. Mol. Biol. 2011, 405, 1027. (36) Zelensky, A. N.; Gready, J. E. FEBS J. 2005, 272, 6179. (37) Drickamer, K.; Taylor, M. E. Curr. Opin. Struct. Biol. 2015, 34, 26. (38) Peng, J. W.; Wagner, G. Biochemistry 1995, 34, 16733. (39) Bowman, G. R.; Geissler, P. L. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 11681.

although the fact that the allosteric network resides in this motional regime is based on the correlated mobility of amino acid side chains. This suggests a change in correlated mobility rather than a change in dynamics. Finally, on the second time scale, major rearrangements in the form of cis/trans prolyl isomerization occur. The different equilibrium populations observed in other CLRs might entail regulatory functions complementing our picture of a receptor architecture that is exquisitely adapted to its function as a PRR, promoting antigen uptake and release.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.6b05458. Detailed experimental methods; supplementary Figures S1−S18 and Tables S1−S9 (PDF)



AUTHOR INFORMATION

Corresponding Authors

*[email protected] *[email protected] Present Address #

M. Ballaschk: Max Delbrück Center for Molecular Medicine, Robert-Rössle-Str. 10, 13092 Berlin, Germany Author Contributions ⊥

J.H. and S.A. contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS C.R. thanks the German Research Society (DFG) for an Emmy Noether fellowship (1944/2-1). B.G.K., S.A., and J.H. acknowledge the DFG collaborative research center SFB 765 for generous support. B.G.K., C.R., and J.H. also thank the Funds of the Chemical Industry (FCI). S.A. and B.G.K acknowledge the computer facilities of the Freie Universität Berlin (ZEDAT) for computer time. C.R. and J.H. further thank the Max Planck Society. J.H. and C.R. wish to thank Prof. Peter H. Seeberger for fruitful scientific discussions and Olaf Niemeyer for technical support. J.H. also thanks Daniel Tesolin and Jia Hui Li for their help in optimization of protein production.



REFERENCES

(1) Osorio, F.; Reis e Sousa, C. Immunity 2011, 34, 651. (2) Forni, D.; Cagliani, R.; Tresoldi, C.; Pozzoli, U.; De Gioia, L.; Filippi, G.; Riva, S.; Menozzi, G.; Colleoni, M.; Biasin, M.; Lo Caputo, S.; Mazzotta, F.; Comi, G. P.; Bresolin, N.; Clerici, M.; Sironi, M. PLoS Genet. 2014, 10, e1004189. (3) Guo, Y.; Feinberg, H.; Conroy, E.; Mitchell, D. A.; Alvarez, R.; Blixt, O.; Taylor, M. E.; Weis, W. I.; Drickamer, K. Nat. Struct. Mol. Biol. 2004, 11, 591. (4) Gramberg, T.; Soilleux, E.; Fisch, T.; Lalor, P. F.; Hofmann, H.; Wheeldon, S.; Cotterill, A.; Wegele, A.; Winkler, T.; Adams, D. H.; Pohlmann, S. Virology 2008, 373, 189. (5) Wragg, S.; Drickamer, K. J. Biol. Chem. 1999, 274, 35400. (6) Snyder, G. A.; Ford, J.; Torabi-Parizi, P.; Arthos, J. A.; Schuck, P.; Colonna, M.; Sun, P. D. J. Virol. 2005, 79, 4589. (7) Tabarani, G.; Thepaut, M.; Stroebel, D.; Ebel, C.; Vives, C.; Vachette, P.; Durand, D.; Fieschi, F. J. Biol. Chem. 2009, 284, 21229. (8) Onizuka, T.; Shimizu, H.; Moriwaki, Y.; Nakano, T.; Kanai, S.; Shimada, I.; Takahashi, H. FEBS J. 2012, 279, 2645. J

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society (40) Keller, B.; Gattin, Z.; van Gunsteren, W. F. Proteins: Struct., Funct., Genet. 2010, 78, 1677. (41) Boehr, D. D.; Schnell, J. R.; McElheny, D.; Bae, S. H.; Duggan, B. M.; Benkovic, S. J.; Dyson, H. J.; Wright, P. E. Biochemistry 2013, 52, 4605. (42) Selvaratnam, R.; VanSchouwen, B.; Fogolari, F.; Mazhab-Jafari, M. T.; Das, R.; Melacini, G. Biophys. J. 2012, 102, 630. (43) Coombs, P. J.; Harrison, R.; Pemberton, S.; Quintero-Martinez, A.; Parry, S.; Haslam, S. M.; Dell, A.; Taylor, M. E.; Drickamer, K. J. Mol. Biol. 2010, 396, 685. (44) Gerasimenko, J. V.; Tepikin, A. V.; Petersen, O. H.; Gerasimenko, O. V. Curr. Biol. 1998, 8, 1335. (45) Yamazaki, T.; Nicholson, L. K.; Wingfield, P.; Stahl, S. J.; Kaufman, J. D.; Eyermann, C. J.; Hodge, C. N.; Lam, P. Y. S.; Torchia, D. A. J. Am. Chem. Soc. 1994, 116, 10791. (46) Kim, M. O.; Blachly, P. G.; McCammon, J. A. PLoS Comput. Biol. 2015, 11, e1004341. (47) Diehl, C.; Engstrom, O.; Delaine, T.; Hakansson, M.; Genheden, S.; Modig, K.; Leffler, H.; Ryde, U.; Nilsson, U. J.; Akke, M. J. Am. Chem. Soc. 2010, 132, 14577. (48) MacRaild, C. A.; Daranas, A. H.; Bronowska, A.; Homans, S. W. J. Mol. Biol. 2007, 368, 822. (49) Nesmelova, I. V.; Ermakova, E.; Daragan, V. A.; Pang, M.; Menendez, M.; Lagartera, L.; Solis, D.; Baum, L. G.; Mayo, K. H. J. Mol. Biol. 2010, 397, 1209.

K

DOI: 10.1021/jacs.6b05458 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX